Priors and Posteriors are almost the same

pnovak · August 29, 2017, 7:40am

I have now estimated a model, where I get the following results:

parameters
prior mean post. mean 90% HPD interval prior pstdev

h 0.740 sigma_c 1.353 sigmal 2.400 epsilonp 0.908 epsilonw 0.737 indp 0.469 indw 0.763 phi_p 1.500 phi_w 1.500 lambda 0.300 phi 1.000 phi_pi 1.500 phi_y 0.500 rhoii 0.750 eta 1.200 rhob 0.750 rhoa 0.750 rhog 0.750 rhoinv 0.750 rhop 0.750 rhow 0.750 rhoi 0.750 rhoyf 0.750 rhopif 0.750 0.7412 0.7412 0.7413 norm 0.0100
1.3589 1.3588 1.3591 norm 0.0500
2.3953 2.3952 2.3954 norm 0.0500
0.9043 0.9041 0.9045 beta 0.0500
0.7433 0.7430 0.7435 beta 0.0500
0.4615 0.4612 0.4617 beta 0.0500
0.7569 0.7566 0.7571 beta 0.0500
1.4977 1.4975 1.4979 norm 0.0500
1.5054 1.5052 1.5057 norm 0.0500
0.3014 0.3014 0.3015 beta 0.0100
1.0068 1.0066 1.0070 norm 0.0500
1.4990 1.4975 1.5004 norm 0.5000
0.5004 0.4999 0.5007 norm 0.1000
0.7351 0.7346 0.7355 beta 0.1000
1.2016 1.2014 1.2017 norm 0.1000
0.7532 0.7529 0.7534 beta 0.1000
0.7493 0.7487 0.7497 beta 0.1000
0.7548 0.7543 0.7552 beta 0.1000
0.7391 0.7388 0.7394 beta 0.1000
0.7515 0.7513 0.7517 beta 0.1000
0.7575 0.7571 0.7577 beta 0.1000
0.7428 0.7426 0.7431 beta 0.1000
0.7436 0.7429 0.7441 beta 0.1000
0.7436 0.7432 0.7439 beta 0.1000

There is almost no difference between priors and posteriors and the posterior distribution is almost identical with the posterior mean. Is this a consequence from my assumption that the priors have a standard deviation of 0.05?

jpfeifer · August 29, 2017, 7:51am

Did you check identification? You should be able to see this already in the mode_check plots as a horizontal likelihood.

pnovak · August 29, 2017, 9:21am

The likelihood is horizontal for a few parameters, ie. three or four, but not for others.

pnovak · August 30, 2017, 7:01am

Now I get the following:

ESTIMATION_CHECKS: There was an error in computing the likelihood for initial parameter values.
ESTIMATION_CHECKS: If this is not a problem with the setting of options (check the error message below),
ESTIMATION_CHECKS: you should try using the calibrated version of the model as starting values. To do
ESTIMATION_CHECKS: this, add an empty estimated_params_init-block with use_calibration option immediately before the estima
tion
ESTIMATION_CHECKS: command (and after the estimated_params-block so that it does not get overwritten):

The variance of the forecast error remains singular until the end of the sample

jpfeifer · August 30, 2017, 7:05am

Try to eliminate the identification errors.
The error messages suggests stochastic singularity. What did you change?

pnovak · August 30, 2017, 7:07am

Instead of a mat-file with all 7 variables separately, I declared a variable ts=dseries(DSGE_data.mat). This generated a dseries object, which was then used for estimation. But this time the estimation does not run and gives this error message.

jpfeifer · August 30, 2017, 7:09am

That is strange. Please provide the codes to replicate the issue.

pnovak · August 30, 2017, 9:20am

Please find attached the mod-file and the dataset:

DSGE_series_dat.mat (6.6 KB)
modelestim.mod (3.9 KB)